Cost-Sensitive Extensions for Global Model Trees: Application in Loan Charge-Off Forecasting Marcin Czajkowski 1 , Monika Czerwonka 2 , and Marek Kretowski 1 1 Faculty of Computer Science, Bialystok University of Technology, Wiejska 45a, 15-351 Bialystok, Poland {m.czajkowski,m.kretowski}@pb.edu.pl 2 Collegium of Management and Finance, Warsaw School of Economics, Al. Niepodleglosci 162, 02-554 Warsaw, Poland monika.czerwonka@sgh.waw.pl Abstract. Most of regression learning methods aim to reduce various metrics of prediction errors. However, in many real-life applications it is prediction cost, which should be minimized as the under-prediction and over-prediction errors have different consequences. In this paper, we show how to extend the evolutionary algorithm (EA) for global induction of model trees to achieve a cost-sensitive learner. We propose a new fitness function which allows minimization of the average misprediction cost and two specialized memetic operators that search for cost-sensitive regres- sion models in the tree leaves. Experimental validation was performed with bank loan charge-off forecasting data which has asymmetric costs. Results show that Global Model Trees with the proposed extensions are able to effectively induce cost-sensitive model trees with average mispre- diction cost significantly lower than in popular post-hoc tuning methods. Keywords: cost-sensitive regression, asymmetric costs, evolutionary al- gorithms, model trees, loan charge-off forecasting. 1 Introduction In the vast number of contemporary systems, information including business, research and medical issues is collected and processed. In real-life data mining problems, the traditional minimization of prediction errors may not be the most adequate scenario. For example, in medical domain misclassifying an ill patient as a healthy one is usually much more harmful than treating a healthy patient as an ill one and sending him for additional examinations. In finance, investors tend to sell winning stocks more readily than losing stocks in the sense that they realize gains relatively more frequently than losses. The sadness that one experiences in losing the money appears to be greater than the pleasure of gaining the same amount of money. This strong loss aversion was explained and described in the prospect theory by Kahneman and Tversky [14] and applied to finance practice by Shefrin and Statman [25]. J. ´ Swi  atek et al. (eds.), Advances in Systems Science, 315 Advances in Intelligent Systems and Computing 240, DOI: 10.1007/978-3-319-01857-7_ 30, c  Springer International Publishing Switzerland 2014